| Computer aided breast mass detection is a key step of computer aided diagnosis of early-stage breast cancer.The traditional computer-aided breast mass detection methods have many deficiencies when it comes to detection accuracy and efficiency.The breast mass detection methods implementation through content-based image retrieval can not only provide detection result,but also return a series of images which are similarity with the query mammogram to doctors.By this way,the detection result will be more credible.In this paper,the research work was carried out on breast mass detection methods realizedby image retrieval and proposed appropriate solutions corresponding flaws.The main works of this thesis and innovations are summarized as follows:Firstly,we overviewed the background,significance and the situation of computer-aided breast mass detection,then gave a classified summary of existing methods of computer-aided breast mass detection,and described several classic related methods in this field.Hashing-based large-scale image retrieval was mainly studied in computer-aided breast mass detection.Secondly,a large-scale image retrieval algorithm based on location preserving iterative quantization has been applied to the computer-aided breast mass detection system.This method can transform the original image features into binary codes which preserved the similarity between original features.Then image retrieval was executed in hamming space.The method solved the problem of slow detection speed and large storage cost when the image number of reference database is large-scale.What is more,the detection efficiency was improved significantly.Lastly,in order to improve the accuracy of breast mass detection,we proposed to better express and distinguish mammograms by fusing local features and global features.The local feature and global feature used in this paper are SIFT and GIST,respectively.Experiment results show that this method can improve the accuracy of mass detection significantly. |
PDF Full Text Request